How practical and digital combined to make that crazy episode 5 ‘Last of Us’ clicker battle.
In ‘Endure and Survive’, the fifth episode of the streaming series The Last of Us, Joel (Pedro Pascal) and Ellie (Bella Ramsey) encounter a resistance movement and an unexpected infestation of the infected, or clickers, just outside Kansas City, Missouri. Not only that, the horde of clickers also includes a heavily mutated and fungal-covered host, or bloater.
A host of live-action performers wearing complex prosthetics and make-up effects designed by BGFX would form the base of the attack. Wētā FX then took scans of the performers, including the bloater and a child clicker, and extensive on-set data to enhance and recreate many of the performances completely in CG.
befores & afters sat down with Wētā FX visual effects supervisor Simon Jung and animation supervisor Dennis Yoo, who worked with production visual effects supervisor Alex Wang, to find out more about the significant digital work on this attack sequence.
How the on-set shoot worked
Both Jung and Yoo had the opportunity to be on set for the live-action shoot. A large contingent of stunt performers, made up in clicker prosthetics and make-up effects, were on location for the extensive scene. Since it was initially relatively undetermined how much CG and digital augmentation work would be needed, the on-set VFX team and Wētā FX carried out as much data acquisition as possible.
“They took HDRIs and there was a photogrammetry booth on site there,” describes Jung. “They were capturing every single ‘infected’ and ran scans for all of us to use later. We also focused mainly on Lidar and witness cameras. That was the main thing we wanted to capture, because we knew how crucial it is for match moves to have good data triangulation.”
“For example,” adds Jung, “our on-set TD Jake Cenac captured a giant master scan of the entire cul-de-sac area. This proved to be really valuable because as it turned out, one of the last shots in the sequence was one we had to go all-CG with. That data was invaluable because we basically had a complete representation of the set and could even get a jump start on postvis with that scan.”
“In that particular shot,” says Yoo, “the camera starts off with the child clicker eating Kathleen (Melanie Lynskey) and then it rises up and cranes and pans over the area. To create that camera move, we needed everything to be CG.”
Cenac also oversaw a fauxcap shoot with several performers after the main photography was completed. “We ran multiple witness cameras all sync’d up and had some of the stunt people in suits to capture data and help us with some of the beats of, for example, the bloater coming out of the sinkhole,” notes Jung.
Ultimately the capture data would not be used in final clicker animation, but the witness cameras did prove useful for reference of the performers. Later, a further motion capture shoot in Wellington, New Zealand was undertaken to give Wētā FX even more performance to work with.
Clickers: finding the right movement and behavior
After an initial edit was crafted for the attack by showrunner Craig Mazin and his team, Wētā FX was initially tasked with both augmenting performers and replacing or adding in completely CG ones. This began with a postvis effort to help shape where additions would be made.
“For this,” outlines Yoo, “we went on our stage and we mimicked what we saw on set. But in the end, those weren’t the motions that Craig wanted. Actually, one of the motions that he particularly called out was this really decrepit, slow-moving actor, which was me in the mocap suit! I think I was sore at the time, so during the capture it actually looks like I’m injured.”
The postvis was refined with further stunt performer motion capture, with particular emphasis given to more erratic clicker behaviors. “From there,” says Yoo, “we actually sped it up. That first shot where you see the hordes of them running out of the pit, we sped our own motion up about 8% just to give it that unearthly feel. We got them running full tilt and acting all crazy, but then we also sped it up 8%.”
This process also helped lock down the final approach to the visual effects. Initially the idea was to digitally augment or replace clicker heads while retaining the performer’s body, however due to the complexity of the match move, compositing, and rotoscoping challenges, this augmentation was not something undertaken as much as first intended.
“We ended up keeping the live action people in there and we would just bulk them up in the areas where it felt a little bit empty or where the performance felt a little bit slow,” observes Jung. “We actually ended up not really augmenting at all. When we needed to, we replaced or added the scene, and because our assets were built to a level that held up pretty well, they brought us a lot of liberty to do that whenever we needed to. We could then rely on Dennis’ capture data and the team of animators to place them where they were needed.”
Yoo praises the work of Wētā FX’s motion crew, motion editors, and animators in making all that possible. “We needed to keep those clickers as real as possible, but we also needed to tweak a lot of the performances. The lighting and comp team did such an amazing job – so much so that once things were starting to get rendered, we’d get these comments from the clients asking for performance changes. Sometimes they would call out their own stuntie from on set and we’d have to say, ‘That’s not us. That’s in the plate.’”
Building the bloater: making it menacing
One of the main CG clicker assets Wētā FX made, of course, was for the bloater. This was based on the extensive BGFX (Barrie and Sarah Gower) make-up effects applied to stunt performer Adam Basil, who performed the role on set in the full suit.
“In any shot that the bloater was in, Adam would be filmed,” outlines Jung. “To be safe, we also captured some fauxcap data in case we were going to have to replace him. The plan initially was to not replace the bloater at all, but to stick with the prosthetics approach. It looked amazing, but I think there’s probably just a limit to how you can perform with a massive suit like that on while still getting the weight, speed, and athleticism you need.”
“No matter how well that suit was made, it was still made out of this foam latex rubber stuff that also moved that way,” adds Yoo. “He moved like he was in a rubber suit and unfortunately all that motion was coming through. He was also a big man, but he wasn’t big enough! In the end, we made him a full head higher.”
The postvis process mentioned earlier was also where Wētā FX established the right look and feel for the motion of the bloater, as Yoo explains. “We had a volume and we’d put little weights on the performer and do different things to make him feel heavier. We’d also enhance it a bit more with keyframing. A lot of it is keyframed, but the base of it is based on a capture.”
One example of where the bloater’s on-set actions were taken to new levels was the initial moment he exits the sinkhole. While the camera move from the live-action plate remained, the scene was ‘Frankenstein’d’ with a few different plates to allow Wētā FX to insert a new CG bloater performance. “Doing that totally helped us nail the performance of the bloater revealing itself,” says Yoo.
What also helped was original bloater reference, research, and a behind-the-scenes test undertaken by Naughty Dog from the video game, something showrunner Neil Druckmann (the creator of the game) provided directly to the VFX studio. “It was pretty cool to see something that was pretty exclusive just to us,” says Jung.
The complex practical suit–with all its intricate fungal growth pieces–was an incredibly useful piece of lighting reference for the visual effects team as it built a CG replica. “It was perfect reference,” advises Jung. “We tried to stay as true as possible to the original design while also incorporating new things, so once it was clear that we were freeing up that character, we got texture supervisor Gino Acevedo on board to help with that, who is obviously a legend at Wētā.”
“It was a huge amount of texturing and shading work,” continues Jung. “The sculpt of the bloater is just absolutely beautiful. It’s almost a shame that you don’t see him in daylight or anything like that. He’s always backlit, so you see little of him. But the detail on him, with all those barnacles and big shields of fungus and growths is mind-boggling.”
Artists at Wētā FX implemented several different material properties on the CG bloater. “Most of him we wanted to be slimy, but he also has little hairy bits,” describes Jung. “We had a little groom through his head–what we call the cordyceps fuzz. And then there were those barnacles on him that were harder shelled and a little bit drier.”
The CG bloater was rigged with a muscle, skin, and soft tissue simulation that enabled the plates to shift against each other, and even deal with his large ‘belly’. “It was a big challenge across all the disciplines at Wētā, but everybody loved it because he’s such a cool and iconic character.”
Bloater vs. Perry
At one point in the melee, the bloater confronts the resistance character Perry (Jeffrey Pierce) and rips off his head. After trying a few passes where Perry would be ripped in half at his torso, feedback came back that it should be a head rip.
“Our lead animator Craig Penn did that,” says Yoo. “The hardest part was finding the performance that they were going to like. We started animating a few things for them and then we found that we couldn’t keep key framing the whole action for them. So we ended up step key framing. We’d give them animation on fives and tens stepped, and then we showed them.”
The action occurs further back in the frame, somewhat out of focus. “The focus is on Kathleen running forward towards camera,” notes Jung. “That aspect added a level of complexity to it, to make sure that it reads. It’s a pretty busy background behind him, for example. We had to make sure that the action was lit properly, even though in reality Perry would’ve been a lot more in shadow.”
“There were also two live action guys running across the plate that didn’t help us there,” adds Jung. “We painted a whole bunch of people out. It was a very, very complex task, that shot. But, again, it turned out it’s such an iconic moment and it looks so cool.”
The challenge of a child clicker
Kathleen meets her demise at the hands of a child clicker, who is also seen earlier stalking Ellie inside a vehicle. A performer on set–who could carry out contortionist-like moves–wore a prosthetic head. “The challenge was that you couldn’t make out that it was a child because obviously it was a head on top of a head and it just looked really big and it didn’t really give Craig that childlike thing that he wanted,” states Jung. “The cool thing was they actually got one of the Naughty Dog concept artists to give us a concept art piece and even an art model. From there on, we did our thing on top of that and developed the look and the textures.”
“It took us a long time to get the proportions right and to make it read as a child,” admits Jung. “Even if you shrink everything down, the problem is that 80% of the head is covered with this giant fungal growth. You only have a cheek and a mouth to basically sell that it is a child. That took quite a long time to get the jawline and all the proportions just right.”
Performance-wise, Wētā FX matched their CG child clicker to the incredible on-set actions of the performer. Says Jung: “We wanted to stay true to it because it was so cool. It was another one where initially we wanted to do a head replacement, but the amount of work after the head-scaling, making the connection work with the body and all of that, it was just easier to paint her out completely. We had to build her whole body anyway for the later scene where she jumps on Kathleen and starts mauling her.”
That mauling moment called for a complete replacement of the child clicker stunt performer in order to also convincingly land all the punch hits. “The idea,” observes Jung, “is that it’s actually not so much punching, it’s more like ripping her face in a really chaotic way.”